Cyber Attack Detection Against State Estimation in CPPS: A Data-Driven Approach
Rahat Naz, K. Seethalekshmi, Nitin Anand Shrivastava · 2024
The amalgamation of Intelligent Electronic Devices (IEDs), smart sensors and data communication network in the legacy power systems have given rise to a new class of systems called Cyber Physical Power Systems (CPPS). The added complexity in the inter-networking of CPPS, has opened up a broad attack surface, increasing its vulnerability towards cyber attacks. To prevent any potential harm to the system and preserve the integrity of the estimated states, a data-driven, regression based cyber attack detection technique is proposed. A Machine Learning (ML) based State Estimator (SE) is proposed and the predicted system states are compared with the one estimated by the conventional SE to detect any malicious data. Conventional Feed-Forward Deep Neural Network (FFDNN) regression model is compared with Extreme Gradient Boost (XGBoost) in training the historical CPPS measurement data to demonstrate that the later is computationally faster by over three times. The proposed FDIA detection technique is validated on a modified IEEE-33 bus system with the successful detection of the bad data injection strategized by two different approaches.